On Wed, May 21, 2025 at 04:40:43PM +0200, Sebastian Andrzej Siewior wrote:
> On 2025-05-21 16:24:59 [+0200], Pablo Neira Ayuso wrote:
> > Hi Sebastian,
> Hi Pablo,
>
> > On Mon, May 12, 2025 at 12:28:44PM +0200, Sebastian Andrzej Siewior wrote:
> > [...]
> > > diff --git a/net/ipv6/netfilter/nf_dup_ipv6.c b/net/ipv6/netfilter/nf_dup_ipv6.c
> > > index 0c39c77fe8a8a..b903c62c00c9e 100644
> > > --- a/net/ipv6/netfilter/nf_dup_ipv6.c
> > > +++ b/net/ipv6/netfilter/nf_dup_ipv6.c
> > > @@ -48,7 +48,7 @@ void nf_dup_ipv6(struct net *net, struct sk_buff *skb, unsigned int hooknum,
> > > const struct in6_addr *gw, int oif)
> > > {
> > > local_bh_disable();
> > > - if (this_cpu_read(nf_skb_duplicated))
> > > + if (current->in_nf_duplicate)
> >
> > Netfilter runs from the forwarding path too, where no current process
> > is available.
>
> If you refer to in-softirq with no task running then there is the idle
> task/ swapper which is pointed to by current in this case. There is one
> idle task for each CPU, they don't migrate.
Thanks for explaining.
I am going to place this series in nf-next.
